This function was written to be efficient when transcoding small - * amounts of data at a time. - * In that case it will be more efficient than \Ref{ucnv_toUnicode}. - * When converting large buffers use \Ref{ucnv_toUnicode}.
+ * Convert a codepage buffer into Unicode one character at a time. + * + * Advantage compared to ucnv_toUnicode() or ucnv_toUChars(): + * - Faster for small amounts of data, for most converters, e.g., + * US-ASCII, ISO-8859-1, UTF-8/16/32, and most "normal" charsets. + * (For complex converters, e.g., SCSU, UTF-7 and ISO 2022 variants, + * it uses ucnv_toUnicode() internally.) + * - Convenient. + * + * Limitations compared to ucnv_toUnicode(): + * - Always assumes flush=TRUE. + * This makes ucnv_getNextUChar() unsuitable for "streaming" conversion, + * that is, for where the input is supplied in multiple buffers, + * because ucnv_getNextUChar() will assume the end of the input at the end + * of the first buffer. + * - Does not provide offset output. + * + * It is possible to "mix" ucnv_getNextUChar() and ucnv_toUnicode() because + * ucnv_getNextUChar() uses the current state of the converter + * (unlike ucnv_toUChars() which always resets first). + * However, if ucnv_getNextUChar() is called after ucnv_toUnicode() + * stopped in the middle of a character sequence (with flush=FALSE), + * then ucnv_getNextUChar() will always use the slower ucnv_toUnicode() + * internally until the next character boundary. + * (This is new in ICU 2.6. In earlier releases, ucnv_getNextUChar() had to + * start at a character boundary.) + * + * Instead of using ucnv_getNextUChar(), it is recommended + * to convert using ucnv_toUnicode() or ucnv_toUChars() + * and then iterate over the text using U16_NEXT() or a UCharIterator (uiter.h) + * or a C++ CharacterIterator or similar. + * This allows streaming conversion and offset output, for example. * *Handling of surrogate pairs and supplementary-plane code points:
* There are two different kinds of codepages that provide mappings for surrogate characters: